20 healthy volunteers will undergo an inclusion visit in order to check inclusion and non inclusion criteria.

Then will be performed:

• Electroencephalography
• MRI

Despite the intra-individual variability, the same object or the same sound repeated several times should generate the most similar connectivity graphs Despite the inter-individual variability, analysis of connectivity graphs must also report high similarity indices between individuals on the same stimuli or stimuli sharing the same semantic properties even if subjects are different.

At the stage of conceptual analysis of stimuli, or from 175 ms after the presentation of the image or sound, the analysis of connectivity graphs should reveal strong similarity indices for several different images of the same object (independence to the visual representation); for picture and sound representing the same object (independence to the sensory modality) or two objects belonging to the same semantic category (conceptual similarity, eg: orange, lemon). Indeed, these objects share common characteristics / semantic dimensions (eg mobile vs. stationary or living vs. non-living etc.).

The density (ie: the ratio between the number of links in a given graph on the total possible number of links), the diameter (ie: the longest path in a graph), the average degree (ie: the average number of links connected to each node), the clustering (ie: the density of connections to a group of nodes with the rest of the network) and other parameters will be compared in terms of standard values available in the literature but also with tools that help to calculate the semantic distance between words such as WordNet and many others.

A computer model of neural network populations in which the experimenter knows in advance where the sources are allows him to test the reliability of his methods while reconstructing graphs and judging their similarity. [8] Finally, the quality of the results about the identification of neural cliques (or complete graphs) could be compared to the artificial neural network model we develop elsewhere [1, 9 and 10]. This model indicates that the distribution of neural cliques (in response to density and efficiency problems) follows a simple principle that we should find in the brain. For example, it is reasonable to think there are many cliques (or complete graphs) at the scale of a small brain region while those cliques are rare when considering spatially distant sources. It is an organization called "small-worlds" and which is classical for the neural networks in the cerebral cortex